張斐章孫建平Chang,Fi-JohnChangSuen, Jian-PingJian-PingSuen2009-02-232018-06-292009-02-232018-06-291997-03http://ntur.lib.ntu.edu.tw//handle/246246/139411類神經網路（ANN）為人們在研究生物腦的過程中發現生物腦的特性，並將此特性 運用至現代的科技，進而發展出對非線性系統能夠自我學習、架構的網路。由於其具有柔性 結構與組織簡潔之特性，並且學習精度高、回想速度快，輸出值亦可為連續值，可解決複雜 映射之問題；藉此特性，對水文事件中所具有之高度不確定、不均勻及隨機等特質提供一新 的方法做分析研究，初步結果令人滿意。 本研究所使用的倒傳遞類神經網路演算法（ BP 演算法）為一具輸入、隱藏及輸出之三層神 經網路；將其應用於降雨－逕流過程之模擬與預測，並探討降雨－逕流過程中前期輸入個數 對訓練與預測之影響；其結果顯示：類神經網路具有學習描述如水文事件複雜關係之能力， 並於預測時獲致良好的成效。With the improvement of science, the characters of biological brains were investigated and implemented to modern technology. The artificial neural network (ANN) has been developed through the conceptual of biological brain characters and shown to be capable of self-organization and self-learning to describe non-linear systems. Due to the flexible structure and simplex organization of the ANN, it could approximately simulate any complex continuous input-output mapping. Consequently, the method is used to investigate the hydrological events which have the characters of highly uncertainty, non-uniform, and randomness. In this study, the back-propagation neural network with three learning network layers, i.e. input, hidden, and output, is utilized to forecast the hourly rainfall and to simulate the rainfall-runoff process. Meanwhile the effect of input number of the rainfall-runoff process is also investigate. The results indicate that the neural network is capable to describe the complex hydrological events and has great forecasting efficiency.en-US類神經網路非線性系統降雨-逕流過程ANNNon-linear systemRainfall-runoff processjournal articlehttp://ntur.lib.ntu.edu.tw/bitstream/246246/139411/1/類神經網路及應用於降雨-逕流過程之研究.pdf